Repeated measurements,log likelihood ratio tests and small samples

Various models have previously been proposed for data comprising m repeated measurements on each of N subjects. Log likelihood ratio tests may be used to help choose between possible models, but these tests are based on distributions which in theory apply only asymptotically. With small N , the log likelihood ratio approximation is unreliable, tending to reject the simpler of two models more often than it should. This is shown by reference to three datasets and analogous simulated data. For two of the three datasets, subjects fall into two groups. Log likelihood ratio tests confirm that for each of these two datasets group means over time differ. Tests suggest that group covariance structures also differ.     

A Note on the Variances of the Tests of Kendall,Jonckheere, and Terpstra

Recently, some papers have been appearing that have noticed some mistakes in the calculations in the statistical software in simple and classical tests (Wilcoxon–Mann-Whitney and Log-Rank). This article examines the variance formulas for the Kendall test, the Jonckheere test, and the Terpstra test. We show in this article that these variances (in spite of what is said in the literature and in the statistical software), in the presence of ties and r ≥ 3 groups, are only equal approximately and are exactly equal for two groups and ties as we prove in the Appendix.     

Multi-Sample Likelihood Ratio Tests Based on Bipolar Watson Distributions Defined on the Hypersphere

We derive likelihood ratio tests for the equality of the directional parameters of k bipolar Watson distributions defined on the hypersphere with common concentration parameter. We analyze the power of these tests in the case of two distributions supposing in the alternative hypothesis two directional parameters forming an angle, which varies from 18° to 90°. We also compare the likelihood ratio tests with a high-concentration F-test.     

Bootstrap Testing for Changes in Persistence with Heavy-Tailed Innovations

This article considers the problem of detection for changes in persistence with heavy-tailed innovations. We adopt a ratio type test and derive its null asymptotic distribution which is dependent on the stable index. Then a residual-based bootstrap is proposed when the stable index is unknown. Our procedure requires drawing bootstrap samples of size m < T, T being the size of original sample. We establish the convergence in probability of the bootstrap distribution function assuming that m → ∞ and m/T → 0. A Monte Carlo study has shown that the bootstrap improve the finite sample size and power compared to the asymptotic test, especially for small stable index.     

Hypothesis Testing and Confidence Regions for the Mean Sojourn Time of an M/M/1 Queueing System

This article discusses testing hypotheses and confidence regions with correct levels for the mean sojourn time of an M/M/1 queueing system. The uniformly most powerful unbiased tests for three usual hypothesis testing problems are obtained and the corresponding p values are provided. Based on the duality between hypothesis tests and confidence sets, the uniformly most accurate confidence bounds are derived. A confidence interval with correct level is proposed.     

Barnard Convex Sets

In calculating significance levels for statistical non inferiority tests, the critical regions that satisfy the Barnard convexity condition have a central role. According to a theorem proved by Röhmel and Mansmann (1999 Röhmel , J. , Mansmann , U. ( 1999 ). Unconditional nonasymptotic one sided tests for independent binomial proportions when the interest lies in showing noninferiority and or superiority . Biometr. J. 2 : 149 – 170 .[Crossref], [Web of Science ®] , [Google Scholar]), when the critical regions satisfy this condition, the significance level for non inferiority tests can be calculated much more efficiently. In this study, the sets that fulfil the Barnard convexity condition are called Barnard convex sets, and because of their relevance, we studied their properties independently of the context from which the sets originated. Among several results, we found that Barnard convex sets are a convex geometry and that each Barnard convex set has a unique basis. Also, we provide an algorithm for calculating the Barnard convex hull of any set. Finally, we present some applications of the concept of the Barnard convex hull of a set for non inferiority tests.     

Bayesian Inference in Marshall–Olkin Bivariate Exponential Shared Gamma Frailty Regression Model under Random Censoring

Many analyses in the epidemiological and the prognostic studies and in the studies of event history data require methods that allow for unobserved covariates or “frailties”. We consider the shared frailty model in the framework of parametric proportional hazard model. There are certain assumptions about the distribution of frailty and baseline distribution. The exponential distribution is the commonly used distribution for analyzing lifetime data. In this paper, we consider shared gamma frailty model with bivariate exponential of Marshall and Olkin (1967 Marshall, A.W., Olkin, I. (1967). A multivariate exponential distribution. J. Am. Stat. Assoc. 62:30–44.[Taylor & Francis Online], [Web of Science ®] , [Google Scholar]) distribution as baseline hazard for bivariate survival times. We solve the inferential problem in a Bayesian framework with the help of a comprehensive simulation study and real data example. We fit the model to the real-life bivariate survival data set of diabetic retinopathy data. We introduce Bayesian estimation procedure using Markov Chain Monte Carlo (MCMC) technique to estimate the parameters involved in the proposed model and then compare the true values of the parameters with the estimated values for different sample sizes.     

New invariant and consistent chi-squared type goodness-of-fit tests for multivariate normality and a related comparative simulation study

ABSTRACT

New invariant and consistent goodness-of-fit tests for multivariate normality are introduced. Tests are based on the Karhunen–Loève transformation of a multidimensional sample from a population. A comparison of simulated powers of tests and other well-known tests with respect to some alternatives is given. The simulation study demonstrates that power of the proposed McCull test almost does not depend on the number of grouping cells. The test shows an advantage over other chi-squared type tests. However, averaged over all of the simulated conditions examined in this article, the Anderson–Darling type and the Cramer–von Mises type tests seem to be the best.